diff --git a/gum/arch-arm64/gumarm64writer.c b/gum/arch-arm64/gumarm64writer.c
index 42683f470..219d301db 100644
--- a/gum/arch-arm64/gumarm64writer.c
+++ b/gum/arch-arm64/gumarm64writer.c
@@ -213,6 +213,7 @@ gum_arm64_writer_init (GumArm64Writer * writer,
   writer->ref_count = 1;
   writer->flush_on_destroy = TRUE;
 
+  writer->data_endian = G_BYTE_ORDER;
   writer->target_os = gum_process_get_native_os ();
   writer->ptrauth_support = gum_query_ptrauth_support ();
   writer->sign = gum_sign_code_address;
@@ -1994,13 +1995,29 @@ gum_arm64_writer_commit_literals (GumArm64Writer * self)
 
     for (slot = first_slot; slot != last_slot; slot++)
     {
-      if (GINT64_FROM_LE (*slot) == r->val)
-        break;
+      if (self->data_endian == G_LITTLE_ENDIAN)
+      {
+        if (GINT64_FROM_LE (*slot) == r->val)
+          break;
+      }
+      else
+      {
+        if (GINT64_FROM_BE (*slot) == r->val)
+          break;
+      }
+
     }
 
     if (slot == last_slot)
     {
-      *slot = GINT64_TO_LE (r->val);
+      if (self->data_endian == G_LITTLE_ENDIAN)
+      {
+        *slot = GINT64_TO_LE (r->val);
+      }
+      else
+      {
+        *slot = GINT64_TO_BE (r->val);
+      }
       last_slot = slot + 1;
     }
 
@@ -2026,13 +2043,28 @@ gum_arm64_writer_commit_literals (GumArm64Writer * self)
 
     for (slot = first_slot; slot != last_slot; slot++)
     {
-      if (GINT32_FROM_LE (*slot) == r->val)
-        break;
+      if (self->data_endian == G_LITTLE_ENDIAN)
+      {
+        if (GINT32_FROM_LE (*slot) == r->val)
+          break;
+      }
+      else
+      {
+        if (GINT32_FROM_BE (*slot) == r->val)
+          break;
+      }
     }
 
     if (slot == last_slot)
     {
-      *slot = GINT32_TO_LE (r->val);
+      if (self->data_endian == G_LITTLE_ENDIAN)
+      {
+        *slot = GINT32_TO_LE (r->val);
+      }
+      else
+      {
+        *slot = GINT32_TO_BE (r->val);
+      }
       last_slot = slot + 1;
     }
 
diff --git a/gum/arch-arm64/gumarm64writer.h b/gum/arch-arm64/gumarm64writer.h
index 9aa8ac01e..1597ee32d 100644
--- a/gum/arch-arm64/gumarm64writer.h
+++ b/gum/arch-arm64/gumarm64writer.h
@@ -34,11 +34,26 @@ G_BEGIN_DECLS
 typedef struct _GumArm64Writer GumArm64Writer;
 typedef guint GumArm64IndexMode;
 
+/*
+ * Valid values for this field are:
+ * - G_LITTLE_ENDIAN
+ * - G_BIG_ENDIAN
+ * - G_BYTE_ORDER (an alias for one of the above)
+ */
+typedef guint GumArm64DataEndian;
+
 struct _GumArm64Writer
 {
   volatile gint ref_count;
   gboolean flush_on_destroy;
 
+  /*
+   * Whilst instructions in AArch64 are always in little endian (even on
+   * big-endian systems), the data is in native endian. Thus since we wish to
+   * support writing code for big-endian systems on little-endian targets and
+   * vice versa, we need to check the writer configuration before writing data.
+   */
+  GumArm64DataEndian data_endian;
   GumOS target_os;
   GumPtrauthSupport ptrauth_support;
   GumAddress (* sign) (GumAddress value);
diff --git a/gum/gumdefs.h b/gum/gumdefs.h
index afec756d8..d1beb686d 100644
--- a/gum/gumdefs.h
+++ b/gum/gumdefs.h
@@ -119,9 +119,12 @@ typedef GumMipsCpuContext GumCpuContext;
  * The only non-legacy big-endian configuration on 32-bit ARM systems is BE8.
  * In this configuration, whilst the data is in big-endian, the code stream is
  * still in little-endian. Since Capstone is disassembling the code stream, it
- * should work in little-endian even on BE8 systems.
+ * should work in little-endian even on BE8 systems. On big-endian 64-bit ARM
+ * systems, the code stream is likewise in little-endian.
  */
-#if G_BYTE_ORDER == G_LITTLE_ENDIAN || defined (__arm__)
+#if G_BYTE_ORDER == G_LITTLE_ENDIAN || \
+    defined (__arm__) || \
+    defined (__aarch64__)
 # define GUM_DEFAULT_CS_ENDIAN CS_MODE_LITTLE_ENDIAN
 #else
 # define GUM_DEFAULT_CS_ENDIAN CS_MODE_BIG_ENDIAN